I think that you need to download sentences for each language in separate files, then download connections file and after that you need to use some graph tools to connect sentences (example NetworkX for Python).

Note first of all that you are talking about a language with more than 777,000 sentences, and working with such a large set of sentences is going to require scripting/programming not just for the downloading of the sentences, and not just for the selection of the subset you want, but also for the management of those sentences on your side.

The download page lets you download the following with the click of a button:
- all sentences in language A with translations in language B
- all sentences in language A
- all sentences in language A that have audio (but not the audio itself, which can only be downloaded if it the license says so, and needs to be downloaded via a URL; this is explained on the downloads page)

Another alternative, as the page says, is to produce a list of sentences, which can then be downloaded. However, this is impractical with the number of sentences you'd be dealing with.

Without scripting/programming knowledge, you could do three downloads, consisting of all Kabyle sentences translated into French, English, and Spanish, respectively. This would give you TSV files (tab-separated rather than comma-separated) containing the sentences and translations. (For reference, the one for French would be 17.8 MB in size and contain more than 200,000 entries.) This is not what you asked for, but it's probably the best you can do without scripting/programming. Otherwise, the help you need is most likely going to go beyond what can be provided on the Wall.